The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides curl command templates for interacting with the vendor's API at https://api.pruna.ai/v1/predictions to generate images.
[EXTERNAL_DOWNLOADS]: The documentation instructs users to install dependencies from the vendor's repository (PrunaAI/pruna-skills) using npx skills add. These references target the official resources of the skill's author.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it ingests untrusted user input for image descriptions and interpolates it into API requests.
Ingestion points: User-provided descriptions for subjects, products, or scenes in SKILL.md instructions.
Boundary markers: The skill mandates a "Fidelity check" and "Lock brief" phase to ensure the final prompt matches user requirements before execution.
Capability inventory: Performs network operations via curl to an external API.
Sanitization: Relies on the agent's adherence to "Prompt craft" rules to reformulate raw input into structured, concrete language.
[CREDENTIALS_UNSAFE]: References the PRUNA_API_KEY using environment variable syntax (${PRUNA_API_KEY}). This is a standard practice for secure secret handling and does not involve hardcoded credentials.

p-image